Eco-efficiency is an important indicator for measuring sustainable development. Clarifying the spatial distribution of eco-efficiency and its evolution is of vital importance to achieving coordinated regional development of China's economy and environment. However, previous research has failed to consider the spatial interaction distribution of eco-efficiency. The SBM model considering undesired output was used to measure the eco-efficiency of 30 provinces (municipalities and autonomous regions) in China from 2005 to 2015, and the spatial econometric model was used to investigate the spatial interaction pattern of eco-efficiency in China. The results show that: (1) From the perspective of the whole country or by region, China's eco-efficiency has generally shown a downward trend. (2) From the perspective of the evolution of the spatial distribution of eco-efficiency, the characteristics of high eco-efficiency distributed in developed provinces have been continuously strengthened, and low eco-efficiency areas are mainly distributed in developing provinces. Beijing, Tianjin, Shandong, Jiangsu, Shanghai, Zhejiang, Fujian, Guangdong, Hainan, Qinghai have always been in the first echelon of eco-efficiency, while Xinjiang, Guangxi and Hunan have always been in the fourth echelon. From 2005 to 2015, the members of each echelon changed, but the distribution characteristics based on the level of economic development remained basically unchanged. (3) At the national level, the estimation results based on the economic weight matrix show that there is a positive spatial interaction effect of eco-efficiency among regions. The improvement of eco-efficiency in a certain region will have a positive spillover effect on its economic neighbors. Extensive learning sharing and competition among different regions may explain the positive eco-efficiency interactions among regions. (4) The sub-sample test shows that the eco-efficiency of developed and developing regions shows a positive spatial interaction effect, and the spatial interaction effect is greater than the spatial interaction effect between the two types of regions, indicating that the distribution in China's eco-efficiency is based on the level of economic development, thus showing the "sorting" based on the level of economic development. This conclusion is still robust even with the geographic distance and adjacent weight matrix.